ranks documents that are news stories, that appeared after 1999, and that contained
at least one text segment of length 20 that mentioned a person named
“brown”, a company name, and at least one of the three words dealing with money.
The inference network model can easily deal with the combination of this type of
evidence, but for simplicity, we have not implemented these operators in Galago.
Another part of the inference network model that we do support in the Galago
query language is document priors. Document priors allow the specification of
a prior probability over the documents in a collection. These prior probabilities
influence the rankings by preferring documents with certain characteristics, such
as those that were written recently or are short.